Skip to main content
. 2021 Feb 18;7:e358. doi: 10.7717/peerj-cs.358

Algorithm 1. Build a deep learning model using CXRVN-proposed architecture.

Input ← Image_COVID-19_Set imds
Output ← CXRVN

  1. Begin

  2.  // Preprocessing COVID-19 X-Ray image(s) in imds

  3.  For i=1: length(imds)

  4.   img ← read(imds,i)

  5.   cxr ← isXGray(img)

  6.   img ← resize(cxr,[128, 128])

  7.   save(imds,I,img)

  8.  End for

  9.  // Build CXRVN Structure

  10.  NLayers ← new Layers{}

  11.  NLayers.append(new Input layer)

  12.  NLayers.append(new Convolutional layer)

  13.  NLayers.append(new Normalization layer)

  14.  NLayers.append(new Relu layer)

  15.  NLayers.append(new Pooling layer)

  16.  NLayers.append(new Convolutional layer)

  17.  NLayers.append(new Normalization layer)

  18.  NLayers.append(new Relu layer)

  19.  NLayers.append(new Pooling layer)

  20.  NLayers.append(new Convolutional layer)

  21.  NLayers.append(new Normalization layer)

  22.  NLayers.append(new Relu layer)

  23.  NLayers.append(new Pooling layer)

  24.  NLayers.append(new FeatureConnected layer)

  25.  NLayers.append(new Softmax layer)

  26.  NLayers.append(new Classification layer)

  27.  // Train CXRVN using options

  28.  Options.set(SolverOptimizer ← mini-batch gradient decent with momentum or Adam)

  29.  Options.set(InitialLearnRate ←1e-3)

  30.  Options.set(LearnRateSchedule ← Piecewise)

  31.  Options.set(MiniBatchSize ←32)

  32.  Options.set(LearnRateDropFactor ←0.2)

  33.  Options.set(LearnRateDropPeriod ←5)

  34.  Options.set(Shuffle←Every Epoch)

  35.  Options.set(ValidationFrequency←2)

  36.  Options.set(MaxEpochs ←20)

  37.  CXRVN ← trainNetwork(NLayers, imds, Options)

  38. End